Spectral Evidence for Substrate Availiability Mitra 2020.Pdf

Agricultural and Forest Meteorology 291 (2020) 108062

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology

journal homepage: www.elsevier.com/locate/agrformet

Spectral evidence for substrate availability rather than environmental control of methane emissions from a coastal forested wetland T ⁎ Bhaskar Mitraa,b, , Kevan Minickc, Guofang Miaod, Jean-Christophe Domece, Prajaya Prajapatia, Steve G. McNultyf, Ge Sunf, John S. Kingc, Asko Noormetsa a Department of Ecology and Conservation Biology, Texas A&M University, College Station, TX, United States b Northern Arizona University, School of Informatics, Computing and Cyber Systems, Flagstaﬀ, AZ 86011, United States c Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC, United States d School of Geographical Sciences, Fujian Normal University, Fuzhou, Fujian Province, China e Bordeaux Sciences Agro, UMR INRA-ISPA 1391, 33195, Gradignan, France f Eastern Forest Environmental Threat Assessment Center, USDA Forest Service, Raleigh, NC, United States

ARTICLE INFO ABSTRACT

Keywords: Knowledge of the dynamics of methane (CH4) fluxes across coastal freshwater forested wetlands, such as those Cross-scale analysis found in the southeastern US remains limited. In the current study, we look at the spectral properties of eco- Methane system net CH4 exchange (NEECH4) time series, and its cospectral behavior with key environmental conditions Plant transportation (temperature (Ts5), water table (WTD) and atmospheric pressure (Pa)) and physiological fluxes (photosynthesis Wavelet analysis (GPP), transpiration (LE), sap flux (J )) using data from a natural bottomland hardwood swamp in eastern North Water table depth s Carolina. NEECH4 fluxes were measured over five years (2012 – 2016) that included both wet and dry years. During the growing season, strong cospectral peaks at diurnal scale were detected between CH4 efflux and GPP,

LE and Js. This suggests that the well understood diurnal cycles in the latter processes may affect CH4 production through substrate availability (GPP) and transport (sap flow and LE). The causality between different time series

was established by the magnitude and consistency of phase shifts. The causal effect of Ts5 and Pa were ruled out because despite cospectral peaks with CH4, their phase relationships were inconsistent. The effect of fluctuations

in WTD on CH4 eﬄux at synoptic scale lacked clear indications of causality, possibly due to time lags and

hysteresis. The stronger cospectral peak with ecosystem scale LE rather than Js suggested that the evaporative component of LE contributed equally with plant transpiration. Hence, we conclude that while the emission of

dissolved gases through plants likely takes place, it may not contribute to higher CH4 emissions as has been proposed by aerenchymatous gas transport in sedge wetlands. These ﬁndings can inform future model devel-

opment by (i) highlighting the coupling between vegetation processes and CH4 emissions, and (ii) identifying speciﬁc and non-overlapping timescales for diﬀerent driving factors.

1. Introduction assessment is critical to developing climate mitigation strategies as (i) the coastal wetlands serve as the main defense against hurricanes and − Wetlands sequester nearly 30% of the global soil carbon despite sea level rise, which at 3.4 mm yr 1 is among the fastest in the world limited geographical range (~ 8%) (Song et al., 2015). Wetlands also (https://tidesandcurrents.noaa.gov/sltrends/), and (ii) they represent a −1 contribute significantly to global methane (CH4) emissions, a powerful significant and unique carbon sink, holding as much as 500 t C ha in greenhouse gas with a global warming potential 28 times greater than the soil (Johnson and Kern, 2003). The annual CH4 budget estimates for −1 that of carbon dioxide (CO2) (IPCC, 2013). A 2.5-fold increase in CH4 the world's wetlands range from 92 to 260 Tg yr (Khalil and emissions since the preindustrial times has generated increased scrutiny Rasmussen, 1983; Walter et al., 2001; Reeburgh, 2003; Denman et al., of it's contribution from different sources including wetlands, oceans, 2007; Meng et al., 2012) and remains a big uncertainty from the per- fossil fuels and rice agriculture (Dlugokencky et al., 2011). Such an spective of global CH4 estimates (Saunois et al., 2016; Melton et al.,

Abbreviations: CO2, Carbon dioxide; WT, Continuous wavelet transformation; COI, Cone of Inﬂuence; CWT, Cross-wavelet transformation; DWT, Discrete wavelet transformation; ET, Evapotranspiration; GPP (umol m-2 s-1), Gross primary productivity; PAR (umol m-2 s-1), Photosynthetically active radiation ⁎ Corresponding author. E-mail address: [email protected] (B. Mitra). https://doi.org/10.1016/j.agrformet.2020.108062 Received 2 July 2019; Received in revised form 21 May 2020; Accepted 29 May 2020 0168-1923/ © 2020 Elsevier B.V. All rights reserved. B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062

List of symbols p(xi,yj) Joint probabilities −2 −1 CH4 (umol m s ) Methane −2 −1 Pa (kPa) Atmospheric pressure NEECH4 (umol m s ) Net ecosystem CH4 exchange −2 −1 Ta (°C) Air temperature NEECO2 (umol m s ) Net ecosystem CO2 exchange Pa (kPa) Atmospheric pressure p Probability distribution of measured data p(xi|yj) Conditional probabilities q Probability distribution of modeled data −2 −2 −1 LE (W m ) Ecosystem-scale latent energy Js,(gm s ) Sap ﬂow velocity H(x) Entropy of variable x Ts5 (°C) Soil temperature at 5 cm H(y) Entropy of variable y WTD (cm) Water table depth J(x,y) Joint entropy of x and y

2013; Bastviken et al. 2011). Such uncertainty primarily exists because temperature overlooks well-known phenomena like time-lags and hys- of the lack of information about CH4 exchange across different wetland tereses (Hatala et al., 2012; Moore and Dalva, 1993; Sachs et al., 2008; types as well as insufficient understanding of the drivers of NEECH4 Meijde et al. 2011; Wille et al. 2008). An alternative to this rudimentary (Megonigal and Guenther, 2006; Heimann, 2011; Olefeldt et al., 2013). approach is a scale-dependent analysis of NEECH4 (Sturtevant et al., Improved understanding of the underlying processes and the interac- 2016). Given the number of steps to net CH4 emission (described tion of different drivers are needed to better capture the magnitude and above), decomposing the CH4 signals into low and high-frequency dynamics of ecosystem CH4 exchange. bands may shed light onto CH4 dynamics and its control mechanisms. CH4 exchange in wetlands is a multidimensional process. CH4 pro- This methodology has been successfully applied to identify the multi- duction occurs primarily in the anaerobic regions of the soil scale drivers of CO2 flux across different biomes (Katul et al., 2001; (Megonigal et al., 2004), although production inside plants has also Stoy et al., 2005; Vargas et al., 2010; Mitra et al. 2019). been hypothesized (Covey and Megonigal, 2019). The emission of dis- In the current study, we will analyze the spectral properties of solved CH4 in soil water to the atmosphere can occur through diffusion ecosystem net CH4 exchange time series, and its cospectral behavior or ebullition (Jeffrey et al., 2019), and may be modified by specialized with key environmental conditions (temperature and water table) and plant anatomical structure, vascular architecture and rooting volume physiological processes (photosynthesis, respiration and transpiration)

(Bhullar et al., 2013; Bhullar et al., 2014). The outgassing of CH4-sa- using a 5-year record from a natural bottomland hardwood swamp in turated water may also occur in plant vasculature, along its transport eastern North Carolina. We hypothesize that the functional dependence pathway to leaves (Nesbit et al., 2009). While the transport of dissolved of CH4 exchange on any of these factors would manifest in co-spectral gases in xylem sap has been established (Teskey et al., 2008; Aubrey signatures with them. At the very least, spectral analysis will allow and Teskey 2009), its precise quantification is difficult. The diffusion of partitioning the sources of variability between fast-changing (radiation,

CH4 through the root and stem spongy tissues (aerenchyma) of some vapor pressure deficit, stomatal conductance, photosynthesis), inter- especially wetland-adapted species (e.g. sedges) has been hypothesized mediate or synoptic-scale (atmospheric pressure, soil moisture avail- to result in elevated emissions compared to non-vegetated surfaces, as ability, water table depth), and seasonal processes (seasonal changes in the off-gassing of CH4 through the plant allows it to bypass the mi- radiation, temperature, phenology). The current study focuses on the crobial re-oxygenation in the anaerobic-aerobic interface in the soil variation in CH4 and its drivers at the first two of the three timescales. (Bubier and Moore, 1994; Joabsson et al., 1999). Although some tree species also have aerenchymatous tissues (e.g. baldcypress, water tu- 2. Methods pelo and Atlantic white cedar), its role in ecosystem CH4 emissions is unclear. Finally, vegetation structure and activity may also affect CH 4 2.1. Study site production, as plant-derived carbohydrates may serve as substrates for archaeal methanogens in the soil (Whiting and Chanton, 1993; The study site (35°47′16.32"N; 75°54′13.74"W) is located in Christensen et al., 2003; Long et al., 2010). As plant carbohydrate status Alligator River National Wildlife Refuge on the Albemarle-Pamlico can affect soil CO production (Mitra et al., 2019), presumably through 2 peninsula in Dare County of eastern North Carolina, USA. The site is an effect on microbial substrate availability, it could also be an im- registered as US-NC4 in the Ameriflux database. Established in 1984, portant regulator of archaeal methanogen activity. Alligator River National Wildlife Refuge is characterized by a hetero- Other well-documented physical drivers of CH production include 4 geneous conglomeration of pocosin wetland types (Allen et al., 2011). water table depth (WTD), atmospheric pressure, and temperature. Overstory vegetation across this freshwater forested wetland consist Water table position operates as an ‘on/off’ switch for CH production. 4 primarily of water tupelo (Nyssa aquatica), red maple (Acer rubrum), Water table drawdown aerates deeper soil horizons, altering the bal- swamp tupelo (Nyssa biflora), along with bald cypress (Taxodium dis- ance between methanogenic archaea and methanotrophic bacteria, re- tichum), sweetgum (Liquidambar styraciflua), white cedar ducing net CH efflux. Fluctuations in atmospheric pressure have been 4 (Chamaecyparis thyoides) and loblolly pine (Pinus taeda). The understory shown to facilitate CH release from wetlands, as the dissolution of 4 vegetation consists of fetterbush (Lyonia lucida), bitter gallberry (Ilex gases in water is in part controlled by pressure (Tokida et al., 2007). albra), red bay (Persea borbonia), and sweet bay (Magnolia virginiana). Finally, temperature is often seen as the primary driver of biogeo- The 30-year (1981-2010) mean temperature and precipitation were chemical fluxes as it controls the available energy (Arrhenius, 1889; − 16.9°C and 1270 mm yr 1 (measured at Manteo airport, NC, about 32 Nahlik and Mitsch, 2011; Sachs et al., 2010), and many models include km from the study site). Canopy height across the site ranged from 15 to empirical temperature response functions for CH production while 4 20 m, with leaf area index peaking at 4.0 ± 0.3 in early July from a water level acts as a simple on/off switch (Olefeldt et al., 2013; Yvon- minimum of 1.3 ± 0.3 during the non-growing season (Domec et al., Durocher et al., 2014; Turetsky, 2014a,b; Vargas et al., 2010; Treat 2015). The pH of soil at the surface varied between 4.2 and 4.8 et al., 2018). − (Minick et al., 2019b). Bulk density of dry soil was 0.08 ± 0.02 g cm 3 Despite the universality of the temperature and water table response (mean ± SD). The primary soil types are poorly drained Pungo and of CH efflux, such simple models oversimplify the system and may 4 Belhaven mucks with their corresponding organic carbon content limit our ability to interpret observed CH flux dynamics. This is due to 4 varying between 40 and 100% and 20 and 100%, respectively (Web Soil the fact that instantaneous combinations of water table and soil Survey; Miao et al., 2017). Precipitation is the primary source of

2 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062 freshwater for this refuge, and astronomical tides are absent due to the Port Orange, FL, USA). All the meteorological variables were measured unique combination of geomorphic features and lagoon environments at 10 sec interval and averaged every 30 minutes. (Moorhead and Brinson, 1995). Other features of this site include the presence of micro-topography, the absence of runoff, and low-intensity 2.4. Sap flow measurements drainage. Granier-type sap flux sensors were used to monitor transpiration at 2.2. Gas flux measurements the individual tree scale (Granier 1985, 1987). Five trees per each predominant species (pond pine, water tupelo, red maple and bald

Ecosystem-scale turbulent fluxes of latent energy (LE), CO2 and CH4 cypress, accounting for more than 75% of the total tree basal area) were exchange were measured with eddy covariance method instrumented near the instrument tower. The 20 mm sensors were in- (Baldocchi et al., 1988). The instrumentation consisted of 3-D sonic stalled on the north side of each tree (at diameter at breast height ~ 1.3 anemometer thermometer (Windmaster, Gill Instruments Limited, m) to reduce the effect of radiation on sap flow estimates. The sensors

Hampshire, UK), an open path CH4 analyzer (LI-7700, LI-COR, Lincoln, were covered with aluminum foil to protect the sap flux sensors from NE, USA), and an enclosed-path CO2 and H2O analyzer (LI-7200, LI- external perturbations, including precipitation, as well as minimize COR). In addition to turbulent fluxes, we also measured CO2 con- temperature fluctuations that would arise from factors other than the centration (LI-820 infrared gas analyzer and a multi-port system) along sap flow. a 5-level profile (~ 0.02, 0.04, 0.3 0.6 and 1 × canopy height) in the The temperature difference data from each sap flux sensor was −2 −1 canopy to estimate canopy storage of CO2, which was added to the converted to sap velocity (Js,gm s )(Ward et al., 2017). Final Js turbulent fluxes of CO2 to quantify net ecosystem exchange of CO2 was the basal-area-weighted sum of Js of the three main canopy species (NEECO2). Detailed information on calculation of CO2 concentration listed above, with the underlying assumption that fluctuations in Js is across our site has already been well documented (Miao et al., 2017). proportional to transpiration dynamics (Small and McConnell, 2008).

Net ecosystem exchange of CH4 (NEECH4) quantified the net CH4 Temperature difference data from each sensor were recorded as half- sequestered or lost by the ecosystem. The sampling frequency of the hourly averages using a CR1000 data logger (Campbell Scientific). raw data was 10 Hz and was recorded using a LI-7550 auxiliary inter- Further information on sapflow-derived tree transpiration at this site face unit (LI-COR, Nebraska, United States). The average canopy height can be found in Domec et al. (2015). at the site was 20 m, and the height of the turbulent flux measurements was 28.2 m in 2012, and 33m since 2013. 2.5. Spectral analysis The turbulent fluxes were calculated and corrections applied with Eddypro software (v 6.1.0, LICOR). The different corrections can be To evaluate the contribution of different environmental and phy- briefly summarized as follows: screening for spikes (Vickers and siological factors with different time constants to the temporal dy-

Mahrt, 1997), rotation of sonic anemometer wind vectors namics of NEECH4, we analyzed the wavelet spectra of NEECH4 and co- (Wilczak et al., 2001), detrending the raw time series (block averaging), spectra of NEECH4 with presumed independent scalar (GPP, LE, Js,Ts5, correction of the time lags between scalar concentration and rotated WTD, Pa) time series in the time-frequency domain. LE and Js were used vertical wind speed, and corrections for variations in air density as the proxies of CH4 transport in tree sap and GPP represented the (Webb et al., 1980). impact of carbon assimilation on NEECH4. Wavelet transformation NEECH4 fluxes measured by the LI-7700 open path analyzer were generated average wavelet spectrum, defined as the average variance corrected for spectroscopic effects along with the density corrections (or energy) associated with specific frequencies for the whole data (Burba et al., 2019). The spectroscopic corrections are required for period. The dominant scale was identified by the maximum of wavelet narrow band laser analyzers to account for the impact of temperature variance (Kumar and Foufoula-Georgiou, 1997). and pressure on the shape of the absorption feature. High (Ibrom et al., Wavelet analysis is advantageous on account of localization in both 2007) and low pass (Moncrieff et al., 2004) filtering corrections were temporal and scale domain, making it robust to deal with non-statio- also applied, and low-quality flux outputs were discarded (Mauder and narities in data. Continous wavelet transformation with the Morlet Foken, 2006). Post-processing of the 30-minute fluxes include filtering (Grinsted et al., 2004) function was used to investigate the oscillation of the data corresponding to low signal strength (<10%) and integral NEECH4 in space and time. The mathematical theory behind this time turbulence characteristics, removal of outliers (Papale et al., 2006) and series strategy has been extensively documented (Torrence and friction velocity thresholding (Goulden et al., 1996). As the tower is Compo, 1998; Grinsted et al., 2004; Stoy et al., 2013) and applied in our surrounded by homogeneous vegetation, there was no directional previous work on soil respiration (Mitra et al., 2019). variability in NEECH4. Final data coverage of NEECH4 and NEECO2 were Apart from average wavelet spectrum, co-spectral or Cross-wavelet 27-47% and 60-75% respectively, depending on a year. NEECO2 (Supp. transformation (CWT) analysis also generated heatmap that provided a Figure S1) was partitioned to gross primary productivity (GPP) and qualitative assessment if one time series was related to another via ecosystem respiration by modeling nighttime NEECO2 as a function of phase diagram. This was accomplished by analyzing regions of high air temperature using a Q10 model. This particular function was used to correlation (red colored areas in heat maps) and at specific frequency predict day and night ecosystem respiration and GPP was quantified as (sub-daily, daily, synoptic) in the heat map. In the heat maps, the the sum of NEECO2 and modeled ecosystem respiration. vertical axis indicated the different frequencies while the horizontal axis referrred to the time of year. The arrows in the heat map only 2.3. Meteorological measurements represent a lag at one certain time and frequency. Direction of arrows in the heat maps provide an understanding of the nature of coupling be-

Micrometeorological variables that were measured above canopy tween NEECH4 and causal scalar (GPP, LE, Js,Ts5, WTD, or Pa) time included air temperature (Ta) and relative humidity (RH; HMP45, series. Arrows in the heat map pointing at right and left would indicate Vaisala, Finland), photosynthetically active radiation (PAR, PARLITE, positive and negative correlation between NEECH4 and drivers,with no Kipp & Zonen, Delft, Netherlands (KZ)), precipitation (TE525, Texas lags. NEECH4 lagged the driver when arrows pointed at up-right (posi- Electronics) and atmospheric pressure (Pa) was measured using a tive correlation) and down-left (negative correlation) direction. Arrows pressure sensor (CS 105, Campbell Scientiﬁc, Logan,UT, USA). pointing at up-left and down-right direction indicate the driver lagging

Belowground measurements include soil temperature at 5 cm (Ts5) and NEECH4. 20 cm depth (Ts20; CS107, Campbell Scientific). Water table depth Causality was inferred only when NEECH4 (i.e. effect) was either in (WTD) was measured with a pressure water level data logger (Infinities, sync or lagged behind the drivers (i.e. the cause) (Banfi and

3 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062

Ferrini, 2012). Another critical aspect identified in the heat map is the scalars preceded NEECH4, the variable with the shortest lag time was cone of influence. Cone of influence is defined as the boundary within identified as the controlling factor, following Koebsch et al. (2015). which wavelet transformation is not affected by edge effects. Trans- formations beyond the cone of influence are ignored. Edge effect arises 2.7. Information theory analysis from the lack of sufficient low frequency data at the beginning and end of the data series. Wavelet multiresolution analysis was combined with information Given the time constant of processes of interest in this study, we theory metrics (Shannon, 1948) to have a better understanding of the constrained the sampling rate for cospectra between 12 hours and 30 mutual dependency between NEE and its drivers across different days. Data were normalized to have zero mean and unit variance and CH4 scales. Multiresolution analysis consisted of wavelet transformation of gaps in the data were padded with zeros. Padding with a constant does an original time series to compute low and high pass filters (d1 – d11). not affect co-spectral power (Mitra et al., 2019) unless the gaps are The Haar basis function (Mahrt, 1991) was used for wavelet transfor- systematically distributed (which was not in our data). We confirmed mation. the absence of spurious spectral peaks by using a null time series of red The information theory metrics of interest was mutual information noise with the same gap structure as the observations, and evaluated content. Mutual information content can be defined as the amount of the WT and CWT spectra and cospectra, similar to Mitra et al. (2019). information that can be gained about variable x when information about another variable y is provided. Mutual information content is fi 2.6. Lag quanti cation greater than zero, with higher values indicative of a larger reduction in uncertainty. Mathematically, mutual information content (MI) can be Instead of the heat maps, we adopted the non-decimated discrete expressed as follows: wavelet transformation (DWT) with Haar basis function (Mahrt, 1991; px y Katul et al., 2001) to quantify lags between NEECH4 and GPP, LE and Js ⎛ i | j ⎞ MIxyHxHyJxy(,)()()(,)Σ·Σ(,)log=+− = pxy ⎜ ⎟ at different scales (Whitcher et al., 2000). For this, the time series of ij ij ⎝ px()i ⎠ (1) NEECH4, GPP, LE and Js was first decomposed into multiple scales (d1 – d11). The different scales (d1-d11) can broadly be classified into Where H(x) and H(y) refers to the entropy of each variable and J(x,y) is diurnal (1 hour – 1.33 days/ d1-d6), synoptic (2.67 - 21.33 days/ d7- the joint entropy, and p(xi,yj) and p(xi|yj) are the joint and conditional d10) and phenological (42.67 days/d11) frequencies. NEECH4 and GPP, probabilities respectively. In this analysis, x was the NEECH4 time series, LE and Js time series at each scale were lagged by ± 12 hours. Typi- and y was each level of high and low band pass filtered versions of cally, Js lags LE due to water storage within trees (Maltese et al., 2018). NEECH4, GPP, LE, Js, Ts5, Pa, and WTD. This allowed us to quantify the This lag time between LE and Js was determined using the above- contribution of each time scale to the predictive uncertainty of NEECH4 mentioned approach and adjusted prior to the calculation of their in- surface exchange. dividual lags against NEECH4. The highest statistically significant (p < As the different measurements had different magnitude and units, 0.05) cross-correlation wavelet coefficient along with the corre- all-time series were normalized to have zero mean and unit variance. sponding lag hours have been reported. If more than one independent This allowed all datasets to be treated equally, before wavelet and cross

Fig. 1. Seasonal variation of 30-minute air temperature (Ta) (Fig. 1A – E), soil temperature at 5 cm depth (Ts5) (Fig. 1F – J), atmospheric pressure (Pa) (Fig. 1K-O), water table depth (WTD) (Fig. 1P-T), gross primary productivity (GPP) (Fig. 1U-Y), latent energy (LE) (Fig. 1Z-A4), sap ﬂux density (Js) (Fig. A5 – A9) and net ecosystem exchange of methane (NEECH4) (Fig. A10-A14) for US-NC4 across diﬀerent years (2012, 2013, 2014, 2015, 2016).

4 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062 analysis. Spectral (“WaveletComp” (Roesch and Schmidbauer, 2014), two time series was inconsistent in 2013 (Supp. Figure S4A), which we “waveslim” (Whitcher, 2015)) and information theory (entropy attribute to greater data gaps in this year. Based on direction of arrow in (Hausser and Strimmer, 2009)) analyses were conducted in R (version the heat map (down-left or left), there were also brief periods when GPP

3.3.2, R Development Core Team, 2016). and NEECH4 were negatively correlated (Supp. Fig. S4A & C). Cospectral analysis between NEECH4 and LE (and Js) yielded similar insight with strong variance at daily scale and weak but significant 3. Results trends at the monthly time step (Fig. 2K-T). Mutual information analysis found the high-frequency band-filtered versions of LE (scales d4- Daily average air temperature (T ) ranged from 5°C in January to a d6) contributing to a greater reduction of uncertainty of NEECH4 (Supp. 28°C in July (Fig. 1A-E). Ta generally remained above 25°C from the Fig. S3 K-O). Mutual information analysis also found a reduction in the middle of June until the end of July. Soil temperature at 5 cm depth uncertainty for NEECH4 by LE at the phenological time scale (d11) (T ) displayed a similar pattern with a lower amplitude (Fig. 1F-J). The s5 (Supp. Figures S3 L-O). Information theory analysis between NEECH4 most pronounced changes in WTD were in response to the annual fall and JS also yielded similar outcomes (results not shown). The magni- storms (Fig. 1P-T). The 5-year study period had normal to above tudes of the diurnal cospectral peaks with NEECH4 were similar, but average precipitation compared to the 30-year normal (Supp. Figure slightly declining in the order of GPP > LE > Js (Fig. 2). S2). NEE (Fig. 1. A10-A14) had a strong seasonal pattern, increasing CH4 Heat maps highlighted strong correlation between NEECH4 and LE at from the beginning of May to July, and declining rapidly from Sep- the diurnal scale (Period (days) = 1) (Fig. 3B,). Both the fluxes oscil- tember to October, where it stayed near annual minimum until April. lated positively with no lags (arrows pointed at right) in 2012, 2015 Although the seasonality patterns of GPP (Fig. 1.U-Y), LE (Fig. 1.Z-A4), and 2016 or NEECH4 lagged (arrows pointed at up-right) LE in 2012 and Js (Fig. 1 A5-A9) were very similar, their cospectra with NEECH4 (Fig. 3B, Supp. Fig. S5C & D). Data gaps were too large in 2013 (Supp. ff ff exhibited distinct di erences, including distinct di erences in the co- Fig. S5A) to establish a consistent relationship. In 2014 and 2015, there spectral peak height (Fig. 2). were instances when LE fluxes lagged NEECH4 (arrows at down-right, Oscillation of NEE was statistically significant at the diurnal CH4 Supp. Fig. S5B & C). LE and NEECH4 were positively coupled at the fi scale, with the presence of weak but signi cant variance also at sy- synoptic scale for brief periods during the growing season in 2012, noptic scales (Fig. 2A-E). Mutual information between CH4 and its 2013, and 2014 (Fig. 3B, Supp. Fig. S5A & B). The heat maps between different band-filtered versions decreased with increase in dyadic time NEECH4 and JS showed similar peaks, but since the peak magnitude was scales (Supp. Fig. S3 A-E). Cospectral analysis between NEECH4 and GPP slightly lower (5-35%) than for LE, the figures are not shown. found significant variance at the diurnal scale, and weak but significant NEECH4- Ts5 cospectra were characterized by significant interactions interactions at the synoptic scale, as well (Fig. 2F-J). The multiscale at the synoptic scale (Fig. 2U-Y), and with small but statistically sig- analysis found a relative reduction in uncertainty in the estimation of nificant cycles also detected at the daily time step in 2013, 2014 & 2015 NEE by GPP at diurnal time scales (~d3-d6) (Supp. Fig. S3 F-J). The CH4 (Fig. 2V-X). MI analysis found the NEECH4-Ts5 interaction to be also only exception was in 2013 (Fig. S3G) when the lowest reduction in strong at the same time scales in all years (Supp. Fig. S3 P-T). The uncertainty was at the phenological time scale (d9-d11). diurnal timestep of Ts5 was also found to significantly reduce un- Heat maps revealed strong correlation between GPP and NEE CH4 certainty in NEECH4 in 2014 and 2015 (Supp. Fig. S3 R & S). Based on during the growing season (May – September) at the diurnal scale direction of arrows in the heat maps, the nature of coupling between Ts5 (Period (days) = 1) (Fig. 3A, Supp. Figures S4). The two signals de- and NEECH4 was inconsistent at the synoptic scale for all years (Fig. 3C, monstrated a strong covariance with no lags (arrow at right) or NEECH4 Supp. Figure S6). slightly lagged GPP (arrow at up-right) in 2012 (Fig. 3A), 2014 (Supp. Pa covaried with NEECH4 at synoptic and monthly frequencies Fig. S4B) and 2016 (Supp. Figure S4D). The relationship between the

Fig. 2. Average wavelet power in the frequency domain generated across diﬀerent years for US-NC4 from the wavelet transformation of net ecosystem exchange of methane (NEECH4) (Fig. A-E); continuous wavelet (CWT) transformation between net ecosystem exchange of methane (NEECH4) and gross primary productivity (GPP)

(Fig. F-J), latent energy (LE) (Fig. K-O), sap ﬂux density (Fig. P-T), soil temperature at 5 cm depth (Ts5) (Fig. U-Y), atmospheric pressure (Pa) (Fig. Z-A4) and water table depth (WTD) (Fig. A5-A9).

5 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062

Fig. 3. Heat maps highlighting for 2012, the cross-wavelet transformation (CWT) between net ecosystem exchange of methane (NEECH4) and diﬀerent drivers including A) gross primary productivity (GPP); B) latent energy (LE); C) soil temperature at 5 cm depth (Ts5); D) atmospheric pressure (Pa) and E) water table depth (WTD). Arrows in the heat map pointing at right and left would indicate positive and negative correlation between NEECH4 and drivers,with no lags. NEECH4 lagged the driver when arrows pointed at up-right (positive correlation) and down-left (negative correlation) direction. Arrows pointing at up-left and down-right direction indicate the driver lagging NEECH4. Red colored areas in heat maps surrounded by white lines indicate areas with 5% signiﬁcance level.

(Fig. 2Z-A4). This was supported by the multiscale analysis, which found MI to increase exponentially, highlighting the importance of low- frequency components of Pa driving the dynamics of NEECH4 (Supp. Fig. S3 U-Y). In the heat maps, areas of high common power between Pa and NEECH4 during the early part of the growing season yielded no consistent relationship from periods 4-16 days across all years (Fig. 3D, Supp. Fig. S7).

Cospectral analysis between NEECH4 and WTD was characterized by signiﬁcant interactions at the synoptic and monthly scale (Fig. 2A5-A9). MI analysis also found the longest time scales of water table dynamics

(d12, Supp. Fig. S3 Z-A4) driving the variation in NEECH4 exchange. In the heat maps, strong correlation between WTD and NEECH4 corresponding to periods greater than four days were primarily concentrated during the late fall season (Supp. Fig. S7C & D). Arrows in the heat maps could not deﬁnitely establish casuaslity as NEECH4 sometimes lagged (arrows pointed at up-right and down-left) and sometimes lead (arrows pointing at up-left) water table ﬂuctuations.

The degree of association for NEECH4 with GPP, LE and Js was quantitatively analyzed for dyadic scale 6 (~ 1.3 days) during the growing season as the strongest variance between the time series were concentrated during that particular phase (Fig. 3A&B, Supp. Fig. S4

&S5). GPP and NEECH4 were positively correlated with the latter (eﬀect) lagging the former (cause) by 0-4.5 hours at scale 6 (Fig. 4A). Js lagged LE by 0.5 to 1.5 hours (2012 – 2016) at dyadic scale 6. After adjusting for this lag time, the lag analysis for NEECH4 with LE (& Js) found the latter leading the former only in 2012 and 2013, with shorter lag for LE in 2013 (Fig. 4B & C). The lags were shorter in both years for LE than for GPP.

4. Discussion Fig. 4. Degree of association (lag/lead) of NEECH4 with GPP (A), LE (B), and Js (C) for dyadic scale 6 (~ 1.3 days). Scalar analysis of CH4 variability has both theoretical and practical

6 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062

implications. By identifying the requisite scale and time, the predictive fold greater in amplitude. Nevertheless, the lower power of the NEECH4 relationship between CH4 and biotic/abiotic drivers can be improved cospectrum with Js than with either GPP or LE (Fig. 2) speaks against with potential to reduce uncertainity in CH4 budget (Yates et al., 2007). the dominance of emissions by transport of dissolved CH4 in the sap. It can also help to develop a comprehensive theoretical framework with Indeed, the phase angles of the NEECH4 and LE also exhibited decou- regard to the drivers of NEECH4. While the full development of these pling at the same time as those of NEECH4 and GPP cospectra (Supp Figs capabilities is beyond the scope of this study, our results oﬀer strong S4A, S4C, S5A, S5C). support to the view that the short-term variability in methane is driven Finally, the evidence for GPP-mediated control of NEECH4 is bol- by substrate availability from photosynthesis, while correlations with stered by the lag time analysis. It has been proposed that the variable environmental conditions appear merely coincidental, resulting from with the shortest lag interval would be the primary driver the same radiation signal that drives GPP. (Koebsch et al., 2015). There is more and stronger evidence for the

NEECH4 exhibited significant variance at sub-daily and diurnal scale substrate availability-based limitation on NEECH4 than for stomatally with weak variance at synoptic, and phenological timescales (Fig 2A-E, controlled release of CH4 with the transpiration stream (although the Supp. Fig. S3A-E). Although the finding of systematic diurnal structure latter is not insignificant). The fact that in some years NEECH4 peaks in NEECH4 is not universal, it is consistent with the reports of Dacey and preceded Js (and LE) suggests that their relationship was correlative, or Klug (1979), Raimbault et al. (1977), Joabsson et al. (1999), confounded by other processes (Supp Figures S5B & S5C). Furthermore,

Hatala et al. (2012), Godwin et al. (2013) and Matthes et al. (2014), the shorter lag time for NEECH4-LE relationship compared to NEECH4-Js and follows the spectral properties reported for NEECO2 (Katul et al., in 2013 (Fig. 4 B & C) suggest that the evaporative component of ET 2001; Braswell et al., 2005; Stoy et al., 2009; Mitra et al., 2019). was equally important to the transport pathway with transpiration. It

Assuming that the power spectrum of response variables (NEECH4 in appears that wherever the evaporation of water occurs, CH4 dissolved this case) is influenced by the power spectra of environmental and in the water gets released, and the trees do not appear to represent a biological forcing, and that a cause precedes effect, we will attribute the preferred low-resistance pathway for CH4 emission (Barba et al., 2019). control of NEECH4 to different environmental factors and biological The relationship of NEECH4 with environmental drivers Ts,Pa and processes based on the power cospectra of these different driving WTD did not lend strong support to widely accepted view of biophysical variables with NEECH4, and the phase lag between the two time series. regulation of CH4 production in wetlands. While these factors covaried For example, the strong cospectral peak between NEECH4 and GPP at the with NEECH4 at diurnal to synoptic time scales, similar to earlier reports daily frequency (Fig. 2F-J, Supp. Fig. S3 F-J) is consistent with earlier (e.g. Tagesson et al., 2012; Hanis et al., 2013; Song et al., 2015; observations across a wide range of wetlands, from subarctic peatlands Klapstein et al., 2014; Tokida et al., 2007), the phase direction in- to subtropical marshes (Whiting and Chanton, 1993; Matthes et al., dicating temporal offsets were either inconsistent (Pa, Ts, WTD) (Supp. 2014). The consistency in the phase direction (arrow at right or up- Figures S6-S8) or negative at diurnal scale (Ts) (Supp. Figure S6C), in- right) between NEECH4 and GPP (Fig. 3A, Supp. Fig. S4), suggest a dicating correlative rather than causal relationship with NEECH4. possible causal connection between the two time series. This hypothesis Large rain events, especially during the hurricane season in the fall, of NEECH4 regulation by carbohydrate substrate supply from GPP is were usually associated with significant changes in NEECH4, and the supported by observations that new photosynthates are the preferred cospectra of NEECH4 and WTD showed strong peaks in 2013, 2015 and carbon substrate for methanogens (Chanton and Whiting, 1996; 2016 (Fig. 2 A5 – A9), but the phase angles (Supp. Fig. S8) failed to Minoda and Kimura, 1994; Hatala et al. 2012; Dorodnikov et al., 2011; indicate consistent causality in this pattern. We therefore conclude that Ström et al., 2012). the effect of water table must have been mediated by some other pro-

The length of the time lag between GPP and NEECH4 (a few hours; cess, possibly plant physiological status. Alternatively, the cospectra Fig. 4A) is consistent with the rate of spread of carbohydrate pressure- may have been obscured by time lags and hysteretic responses. For concentration waves throughout the plant (Thompson and example, the substrate pools and activation times of different microbial Holbrook, 2003), and similar to time lags being observed between GPP populations may cause delayed responses on account of short-lived and soil CO2 efflux in many earlier reports (Mitra et al., 2019; Liu et al., peaks in water table (e.g. Kettunen et al., 1999; Blodau and 2006). It is important to note that studies using tracer analysis in- Moore, 2003; Knorr et al., 2008). It is also possible that there is a variably observe longer time periods for actual mass flow of carbohy- threshold response of NEECH4 to fluctuations in the water table, for drates (Kuzyakov and Gavrichkova, 2010;Mencuccini and Hölttä, 2010; example, as shown for CO2 by Miao et al. (2013). Similarly, Brown et al. Wingate et al., 2010). The mass flow of carbohydrates has a finite and (2014) showed that maximum methanogen activity occurred at a par- − rather uniform speed, usually around 20-30 cm hr 1, and thus the ticular critical water table depth at which the redox potential and delay of assimilation and label detection in target tissues is proportional substrate availability were at the optimum. to the length of the transport pathway, whereas pressure-concentration While the evidence for GPP-mediated control of NEECH4 appears waves alter carbohydrate availability throughout the plant at a shorter strong and holds up to different angles of scrutiny, it could be asked if timescale, and the effect is independent of distances involved omitting CH4 storage in the canopy profile could have affected the (Kuzyakov and Gavrichkova, 2010; Thompson and Holbrook, 2003). findings. As CH4 profile was not measured at the study site, the only There were also periods when GPP and the NEECH4 time series were option for estimating canopy storage was via the single-point rate of either decoupled (Supp. Fig. S4B) or negatively correlated (Supp. Fig. change approach (Hollinger et al., 1994), which in our experience is

S4A & C). This could be indicative of suppressed CH4 production and very unreliable for estimating CO2 storage in tall canopies, including at emission (for example, possibly due to increased CH4 oxidation; the current study site. Nevertheless, the inclusion of this storage esti- Strom et al., 2005; Bouchard et al., 2007), or the dominance of an al- mate did not significantly alter the spectral signature of NEECH4, as well ternative control mechanism over substrate limitation. One such al- as the cospectra with GPP (data not shown). In addition, the magnitude ternative mechanism could be the transport of dissolved CH4 with of the daily cospectral peaks and the lag times remained the similar to xylem sap, that is then released with transpiration, or that could also those without storage. diffuse out through the stem along the way (Barba et al., 2019). This In summary, these findings call for re-evaluation of the common would explain the strong covariance of CH4 at diurnal scale with both correlative patterns as the current way to model CH4 production in LE and Js (Fig. 2K-T, Supp. Fig. S3 K-O). Like GPP, LE and Js are tightly wetlands (e.g. Bridgham et al., 2013; Chu et al., 2014; Turetsky et al., regulated by stomatal conductance, and partitioning their effect on 2014). The spectral analysis reported here has identified certain time

NEECH4 with spectral analysis tools is challenging. The cospectral peak (growing season) and frequency (diurnal) patterns that dominate the between LE and GPP (Supp. Figure S9, only shown for 2012) is very multidimensional interaction between NEECH4 and its biotic and abiotic similar in shape to the cospectrum of NEECH4 and GPP, except about 4- drivers. We posit that the insights of the signiﬁcance of GPP-mediated

7 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062 substrate availability for soil microbial activity (Mitra et al., 2019) have Conversion of natural forests to managed forest plantations decreases tree resistance the potential to stimulate emergent ideas for model development. One to prolonged droughts. For. Ecol. Manag. 355, 58–71. https://doi.org/10.1016/j. foreco.2015.04.012. such approach could be time-varying parameters with specific spectral Dorodnikov, M., Knorr, K.H., Kuzyakov, Y., Wilmking, M., 2011. Plant-mediated CH4 constraints. Second, the role of dissolved CH4 transport in the sap may transport and contribution of photosynthates to methanogenesis at a boreal mire: a C- benefit from the detailed representation of plant hydraulic properties, 14 pulse-labeling study. Biogeosciences 8 (8), 2365–2375. Granier, A., 1985. A new method of sap flow measurement in tree stems. Ann. Des. Sci. and stem water storage that are being introduced to ecosystem models Forest. 42 (2), 193–200. like CLM and Noah-MP (Li & Matheny, personal communication). Granier, A., 1987. Evaluation of transpiration in a douglas-fir stand by means of sap flow measurements. Tree Physiol. 3 (4), 309–319. Acknowledgements Goulden, M.L., Munger, JW., Fan, SM., Daube, B.C., Wofsy, S.C., 1996. Exchange of carbon dioxide by a deciduous forest: Response to interannual climate variability. Science 271, 1576–1578. Funding for this AmeriFlux core site was provided by the U.S. Grinsted, A., Moore, J.C., Jevrejeva, S., 2004. Application of the cross wavelet transform Department of Energy's Office of Science, US DOE National Institute of and wavelet coherence to geophysical time series. Nonlinear Processes Geophys. 11 (5–6), 561–566. Climatic Change Research (08-SC-NICCR-1072), the USDA Forest Hanis, K.L., Tenuta, M., Amiro, B.D., Papakyriakou, T.N., 2013. Seasonal dynamics of Service Eastern Forest Environmental Threat Assessment Center (08-JV- methane emissions from a subarctic fen in the Hudson Bay Lowlands. Biogeosciences – 11330147-038, 13-JV-11330110-081), the US DOE Terrestrial 10 (7), 4465 4479. Hatala, J.A., Detto, M., Baldocchi, D.D., 2012. Gross ecosystem photosynthesis causes a Ecosystems Program (11-DE-SC0006700) and the National Science diurnal pattern in methane emission from rice. Geophys. Res. Lett. 39. Foundation (NSF-IOS 1754893). Maxwell Wightman and Jonathan Heimann, M., 2011. Atmospheric science enigma of the recent methane budget. Nature Furst maintained the site. We thank Dennis Stewart, Scott Lanier and 476 (7359), 157–158. ff Hollinger, D.Y, Kelliher, F.M., Byers, J.N, Hunt, J.E., Mcseveny, T.M, Weir, P.L, 1994. the sta of Alligator River National Wildlife Refuge for site access and Carbon dioxide exchange between an undisturbed old-growth temperate forest and logistical support. Data used for this study can be downloaded from the the atmosphere. Ecology 75 (1), 134–150 1994. Ameriflux database (https://ameriflux.lbl.gov/data/data-availability/). Ibrom, A., Dellwik, E., Flyvbjerg, H., Jensen, N.O., Pilegaard, K., 2007. Strong low-pass filtering effects on water vapour flux measurements with closed-path eddy correlation systems. Agric. For. Meteorol. 147 (3-4), 140–156. Supplementary materials Jeffrey, L.C., Maher, D.T., Johnston, S.G., Kelaher, B.P., Steven, A., Tait, D.R., 2019. Wetland methane emissions dominated by plant-mediated fluxes: Contrasting emis- Supplementary material associated with this article can be found, in sions pathways and seasons within a shallow freshwater subtropical wetland. Limnol. Oceanogr. 0 (0). https://doi.org/10.1002/lno.11158. the online version, at doi:10.1016/j.agrformet.2020.108062. Joabsson, A., Christensen, T.R., Wallen, B., 1999. Vascular plant controls on methane emissions from northern peatforming wetlands. Trends Ecol. Evol. 14 (10), 385–388. fl References Katul, G., et al., 2001. Multiscale analysis of vegetation surface uxes: from seconds to years. Adv. Water Res. 24 (9–10), 1119–1132. Kettunen, A., et al., 1999. Methane production and oxidation potentials in relation to Arrhenius, S., 1889. Über die Reaktionsgeschwindigkeit bei der Inversion von Rohrzucker water table fluctuations in two boreal mires. Soil Biol. Biochem. 31 (12), 1741–1749. durch Sauren. Zeitschrift für Physikalische Chemie 4, 226–248. Khalil, M.A.K., Rasmussen, R.A., 1983. Sources, sinks, and seasonal cycles of atmospheric Baldocchi, D.D., Hicks, B.B., Meyers, T.P., 1988. Measuring biosphere-atmosphere ex- methane. J. Geophys. Res.-Oceans 88 (NC9), 5131–5144. changes of biologically related gases with micrometeorological methods. Ecology 69 Klapstein, S.J., et al., 2014. Controls on methane released through ebullition in peatlands (5), 1331–1340. a ffected by permafrost degradation. J. Geophys. Res.-Biogeosci. 119 (3), 418–431. Barba, J., et al., 2019. Methane emissions from tree stems: a new frontier in the global Knorr, K.H., Oosterwoud, M.R., Blodau, C., 2008. Experimental drought alters rates of soil carbon cycle. New Phytol. 222 (1), 18–28. https://doi.org/10.1111/nph.15582. respiration and methanogenesis but not carbon exchange in soil of a temperate fen. Banfi, F., Ferrini, G., 2012. Wavelet cross-correlation and phase analysis of a free canti- Soil Biol. Biochem. 40 (7), 1781–1791. lever subjected to band excitation. Beilstein J. Nanotechnol. 3, 294–300. Koebsch, F., Jurasinski, G., Koch, M., Hofmann, J., Glatzel, S., 2015. Controls for multi- Bastviken, D., Tranvik, L.J., Downing, J.A., Crill, P.M., Enrich-Prast, A., 2011. Freshwater scale temporal variation in ecosystem methane exchange during the growing season methane emissions offset the continental carbon sink. Science 331 (6013), 50 50. of a permanently inundated fen. Agric. For. Meteorol. 204, 94–105. Bhullar, G.S., Edwards, P.J., Venterink, H.O., 2013. Variation in the plant-mediated Kumar, P., Foufoula-Georgiou, E., 1997. Wavelet analysis for geophysical applications. methane transport and its importance for methane emission from intact wetland peat Rev. Geophys. 35 (4), 385–412. mesocosms. J. Plant Ecol. 6 (4), 298–304. Kuzyakov, Y., Gavrichkova, O., 2010. "REVIEW: time lag between photosynthesis and Bhullar, G.S., Edwards, P.J., Venterink, H.O., 2014. Influence of different plant species on carbon dioxide efflux from soil: a review of mechanisms and controls. Glob. Change methane emissions from soil in a restored Swiss Wetland. PLoS One 9 (2). Biol. 16 (12), 3386–3406. Blodau, C., Moore, T.R., 2003. Experimental response of peatland carbon dynamics to a Liu, Q, Edwards, N.T, Post, W.M, Gu, L, Ledford, J, Lenhart, S, 2006. Temperature-in- water table fluctuation. Aquat. Sci. 65 (1), 47–62. dependent diel variation in soil respiration observed from a temperate deciduous Bouchard, V., Frey, S.D., Gilbert, J.M., Reed, S.E., 2007. Effects of macrophyte functional forest. Glob. Change Biol. 12 (11), 2136–2145. group richness on emergent freshwater wetland functions. Ecology 88 (11), Long, K.D., Flanagan, L.B., Cai, T., 2010. Diurnal and seasonal variation in methane 2903–2914. emissions in a northern Canadian peatland measured by eddy covariance. Glob. Braswell, B.H., Sacks, W.J., Linder, E., Schimel, D.S., 2005. Estimating diurnal to annual Change Biol. 16 (9), 2420–2435. ecosystem parameters by synthesis of a carbon flux model with eddy covariance net Mahrt, L., 1991. Eddy asymmetry in the sheared heated boundary-layer. J. Atmospheric ecosystem exchange observations. Glob. Change Biol. 11 (2), 335–355. Sci. 48 (3), 472–492. Bridgham, S.D., Cadillo-Quiroz, H., Keller, J.K., Zhuang, Q.L., 2013. Methane emissions Maltese, A., Awada, H., Capodici, F., Ciraolo, G., La Loggia, G., Rallo, G, 2018. On the use from wetlands: biogeochemical, microbial, and modeling perspectives from local to of the Eddy covariance latent heat flux and sap flow transpiration for the validation of global scales. Glob. Change Biol. 19 (5), 1325–1346. a surface energy balance model. Remote Sens. 10 (2), 195. 2018. https://doi.org/10. Bubier, J.L., Moore, T.R., 1994. An ecological perspective on methane emissions from 3390/rs10020195. northern wetlands. Trends Ecol. Evol. 9 (12), 460–464. Matthes, J.H., Sturtevant, C., Verfaillie, J., Knox, S., Baldocchi, D., 2014. Parsing the Burba, G, Anderson, A, Komissarov, A, 2019. Accounting for spectroscopic effects in laser- variability in CH4 flux at a spatially heterogeneous wetland: Integrating multiple based open-path eddy covariance flux measurements. Glob. Change Biol. 25 (2019), eddy covariance towers with high-resolution flux footprint analysis. J. Geophys. Res.- 2189–2202. Biogeosci. 119 (7), 1322–1339. Chanton, J.P., Whiting, G.J., 1996. Methane stable isotopic distributions as indicators of Mauder, M., Foken, T., 2006. Impact of post-field data processing on eddy covariance flux gas transport mechanisms in emergent aquatic plants. Aquat. Bot. 54 (2–3), 227–236. estimates and energy balance closure. Meteorol. Z. 15 (6), 597–609. Christensen, T.R., et al., 2003. Factors controlling large scale variations in methane Megonigal, J.P., Hines, M.E., Visscher, P.T., 2004. Anaerobic metabolism: linkages to emissions from wetlands. Geophys. Res. Lett. 30 (7). trace gases and aerobic processes. In: Schlesinger, W.H. (Ed.), Biogeochemistry. Chu, H.S., et al., 2014. Net ecosystem methane and carbon dioxide exchanges in a Lake Elsevier-Pergamon, Oxford, pp. 317–424. Erie coastal marsh and a nearby cropland. J. Geophys. Res.-Biogeosci. 119 (5), Meijide, A., et al., 2011. Seasonal trends and environmental controls of methane emis- 722–740. sions in a rice paddy field in Northern Italy. Biogeosciences 8 (12), 3809–3821. Covey, K.R., Megonigal, J.P., 2019. Methane production and emissions in trees and for- Melton, J.R., et al., 2013. Present state of global wetland extent and wetland methane ests. New Phytol. 222 (1), 35–51. https://doi.org/10.1111/nph.15624. modelling: conclusions from a model inter-comparison project (WETCHIMP). Dacey, J.W.H., Klug, M.J., 1979. Methane efflux from lake-sediments through water lilies. Biogeosciences 10 (2), 753–788. Science 203 (4386), 1253–1255. Mencuccini, M., Holtta, T, 2010. The significance of phloem transport for the speed with Dlugokencky, E.J., Nisbet, E.G., Fisher, R., Lowry, D., 2011. Global atmospheric methane: which canopy photosynthesis and belowground respiration are linked. New Phytol. budget, changes and dangers. Philos. Trans. R. Soc. A-Math. Phys. Eng. Sci. 369 185 (1), 189–203. (1943), 2058–2072. Meng, L., et al., 2012. Sensitivity of wetland methane emissions to model assumptions: Domec, J.C., King, J.S, Ward, E., Oishi, A.C., Palmroth, S., Radecki, A., Bell, D.M., Miao, application and model testing against site observations. Biogeosciences 9 (7), G., Gavazzi, M., Johnson, D.M., McNulty, S.G., Sun, G., Noormets, A., 2015. 2793–2819.

8 B. Mitra, et al. Agricultural and Forest Meteorology 291 (2020) 108062

Miao, G.F., et al., 2017. Hydrology and microtopography control carbon dynamics in cross-biome analysis across multiple time scales. Biogeosciences 6 (10), 2297–2312. wetlands: implications in partitioning ecosystem respiration in a coastal plain Strom, L., Mastepanov, M., Christensen, T.R., 2005. Species-specificeffects of vascular forested wetland. Agric. For. Meteorol. 247, 343–355. plants on carbon turnover and methane emissions from wetlands. Biogeochemistry 75 Miao, G.F., et al., 2013. The effect of water table fluctuation on soil respiration in a lower (1), 65–82. coastal plain forested wetland in the southeastern US. J. Geophys. Res.-Biogeosci. 118 Strom, L., Tagesson, T., Mastepanov, M., Christensen, T.R., 2012. Presence of Eriophorum (4), 1748–1762. scheuchzeri enhances substrate availability and methane emission in an Arctic wet- Minoda, T., Kimura, M., 1994. Contribution of photosynthesized carbon to the methane land. Soil Biol. Biochem. 45, 61–70. emitted from paddy fields. Geophys. Res. Lett. 21 (18), 2007–2010. Sturtevant, C., et al., 2016. Identifying scale-emergent, nonlinear, asynchronous processes Minick, K.J., Li, X., Mitra, B., Noormets, A., King, J.S., 2019b. Water table drawdown of wetland methane exchange. J. Geophys. Res.-Biogeosci. 121 (1), 188–204. alters soil and microbial carbon pool size and isotope composition in coastal fresh- Tagesson, T., et al., 2012. Land-atmosphere exchange of methane from soil thawing to soil water forested wetlands. Front. For. Glob. Change doi: 0.3389/ffgc.2019.00007. freezing in a high-Arctic wet tundra ecosystem. Glob. Change Biol. 18 (6), Mitra, B., Miao, G., Minick, K., Gavazzi, M., Sun, G., McNulty, S., King, J.S., Noormets, A, 1928–1940. 2019. Disentangling the effects of temperature, moisture and substrate availability on Thompson, M.V., Holbrook, N.M., 2003. Scaling phloem transport:Water potential equi-

soil CO2 efflux. JGR-Biogeosci. https://doi.org/10.1029/2019JG005148. librium and osmoregulatory flow. Plant Cell Environ. 26, 1561–1577. Moore, T.R., Dalva, M., 1993. The influence of temperature and water-table position on Tokida, T., et al., 2007. Falling atmospheric pressure as a trigger for methane ebullition carbon-dioxide and methane emissions from laboratory columns of peatland soils. J. from peatland. Glob. Biogeochem. Cycles 21 (2). Soil Sci. 44 (4), 651–664. Tokida, T., et al., 2007. Episodic release of methane bubbles from peatland during spring Moorhead, K.K., Brinson, M.M., 1995. Response of wetlands to rising sea-level in the thaw. Chemosphere 70 (2), 165–171. lower coastal-plain of north-carolina. Ecol. Appl. 5 (1), 261–271. Torrence, C., Compo, G.P., 1998. A practical guide to wavelet analysis. Bull. Am. Nahlik, A.M., Mitsch, W.J., 2011. Methane Emissions from created riverine Wetlands (vol Meteorol. Soc. 79 (1), 61–78. 30, pg 783, 2010). Wetlands 31 (2), 449–450. Turetsky, M.R., et al., 2014. A synthesis of methane emissions from 71 northern, tem- Nisbet, R.E.R., et al., 2009. Emission of methane from plants. Proc. R. Soc. B-Biol. Sci. 276 perate, and subtropical wetlands. Glob. Change Biol. 20 (7), 2183–2197. (1660), 1347–1354. Turetsky, M.R., et al., 2014. A synthesis of methane emissions from 71 northern, tem- Olefeldt, D., Turetsky, M.R., Crill, P.M., McGuire, A.D., 2013. Environmental and physical perate, and subtropical wetlands. Glob. Change Biol. 20 (7), 2183–2197. controls on northern terrestrial methane emissions across permafrost zones. Glob. Vargas, R., Detto, M., Baldocchi, D.D., Allen, M.F., 2010. Multiscale analysis of temporal

Change Biol. 19 (2), 589–603. variability of soil CO2 production as influenced by weather and vegetation. Glob. Papale, D., et al., 2006. Towards a standardized processing of Net Ecosystem Exchange Change Biol. 16 (5), 1589–1605. measured with eddy covariance technique: algorithms and uncertainty estimation. Vickers, D., Mahrt, L., 1997. Quality control and flux sampling problems for tower and Biogeosciences 3 (4), 571–583. aircraft data. J. Atmos. Oceanic Technol. 14 (3), 512–526. Raimbault, M., Rinaudo, G., Garcia, J.L., Boureau, M., 1977. A device to study metabolic Walter, B.P., Heimann, M., Matthews, E., 2001. Modeling modern methane emissions gases in the rice rhizosphere. Soil Biol. Biochem. 9 (3), 193–196. from natural wetlands 1. Model description and results. J. Geophys. Res.-Atmos. 106 Roesch, A., Schmidbauer, R., (2014). WaveletComp: Computational Wavelet Analysis. (D24), 34189–34206. Sachs, T., Giebels, M., Boike, J., Kutzbach, L., 2010. Environmental controls on CH4 Ward, E.J., et al., 2017. TRACC: an open source software for processing sap flux data from emission from polygonal tundra on the microsite scale in the Lena river delta. Siberia thermal dissipation probes. Trees-Struct. Funct. 31 (5), 1737–1742. Glob. Change Biol. 16 (11), 3096–3110. Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements for density Sachs, T., Wille, C., Boike, J., Kutzbach, L., 2008. Environmental controls on ecosystem- effects due to heat and water-vapor transfer. Q. J. R. Meteorolog. Soc. 106 (447),

scale CH4 emission from polygonal tundra in the Lena River Delta. Siberia. J. 85–100. Geophys. Res.-Biogeosci. 113. Whitcher, B., Guttorp, P., Percival, D.B., 2000. Wavelet analysis of covariance with ap- Saunois, M., et al., 2016. The global methane budget 2000-2012. Earth Syst. Sci. Data 8 plication to atmospheric time series. J. Geophys. Res.-Atmos. 105 (D11), (2), 697–751. 14941–14962. Shannon, C.E., 1948. A mathematical theory of communication. Bell Syst. Tech. J. 27 (3), Whiting, G.J., Chanton, J.P., 1993. Primary production control of methane emission from 379–423. wetlands. Nature 364 (6440), 794–795. Small, E.E., McConnell, J.R., 2008. Comparison of soil moisture and meteorological Wilczak, J.M., Oncley, S.P., Stage, S.A., 2001. Sonic anemometer tilt correction algo- controls on pine and spruce transpiration. Ecohydrology 1 (3), 205–214. rithms. Boundary Layer Meteorol. 99 (1), 127–150. Song, W.M., et al., 2015. Methane emissions from an alpine wetland on the Tibetan Wille, C., Kutzbach, L., Sachs, T., Wagner, D., Pfeiﬀer, E.M., 2008. Methane emission from Plateau: neglected but vital contribution of the nongrowing season. J. Geophys. Res.- Siberian arctic polygonal tundra: eddy covariance measurements and modeling. Glob. Biogeosci. 120 (8), 1475–1490. Change Biol. 14 (6), 1395–1408. Stoy, P.C., et al., 2013. Evaluating the agreement between measurements and models of Wingate, L., Ogee, J, Burlett, R, Bosc, A, Devaux, M, Grace, J, Loustau, D, Gessler, A, net ecosystem exchange at diﬀerent times and timescales using wavelet coherence: an 2010. Photosynthetic carbon isotope discrimination and its relationship to the carbon example using data from the North American Carbon Program Site-Level Interim isotope signals of stem, soil and ecosystem respiration. New Phytol. 188 (2), Synthesis. Biogeosciences 10 (11), 6893–6909. 576–589. Stoy, P.C., et al., 2005. Variability in net ecosystem exchange from hourly to inter-annual Yates, T.T., Si, B.C., Farrell, R.E., Pennock, D.J., 2007. Time, location, and scale depen- time scales at adjacent pine and hardwood forests: a wavelet analysis. Tree Physiol 25 dence of soil nitrous oxide emissions, soil water, and temperature using wavelets, (7), 887–902. cross-wavelets, and wavelet coherency analysis. J. Geophys. Res.-Atmos. 112 (D9). Stoy, P.C., et al., 2009. Biosphere-atmosphere exchange of CO2 in relation to climate: a